Collection 6 update: MODIS ‘Deep Blue’ aerosol Andrew M. Sayer, N. Christina Hsu, Corey Bettenhausen, Myeong-Jae Jeong, Jaehwa Lee

MODIS Red-Green-Blue composite MODIS Dark Target AOD MODIS Deep Blue AOD Dust storm: 6 th April 2001

Some applications of Deep Blue data… (1) Global dust source mapping Ginoux et al. (2012) Global-scale attribution of anthropogenic and natural dust sources and their emission rates based on MODIS Deep Blue aerosol products, Rev. Geophys., 50, RG3005, doi: /2012RG

Some applications of Deep Blue data… (2) Comparison of aerosol optical depth from multiple satellite datasets during an intense Saharan dust storm in March 2006 Carboni et al. (2012), Intercomparison of desert dust optical depth from satellite measurements, Atmos. Meas. Tech., 5, , doi: /amt

Some applications of Deep Blue data… (3) Combination with Dark Target aerosol data to compare aerosol/cloud/precipitation relationships between models and satellite observations Yi et al. (2012), Aerosol-cloud-precipitation relationships from satellite observations and global climate model simulations, J. Appl. Remote Sens. 6(1), , doi: /1.JRS

Some applications of Deep Blue data… (4) Constraints on mineral dust aerosol emission in the Gobi and Taklamakan deserts through combination with MISR data and the Geos-Chem model Wang et al. (2012), Top-down estimate of dust emissions through integration of MODIS and MISR aerosol retrievals with the GEOS- Chem adjoint model, Geophys. Res. Lett., 39, L08802, doi: /2012GL

Main developments Collection 6 will include various refinements to Deep Blue: 1.Extended coverage to vegetated surfaces, as well as bright land. 2.Improved surface reflectance models. 3.Improved aerosol microphysical models 4.Improved cloud screening 5.Simplified quality assurance (QA) flags 6.Calibration improvements will mean that Deep Blue can be applied to the whole MODIS record 7.Merged Deep Blue – Dark Target aerosol SDS, to provide a dataset with fewer gaps, for visualisation purposes

Extended spatial coverage

Cloud screening In Collection 5, some cloud-free areas were flagged as cloudy by the 1.38 micron (cirrus/high cloud) test – Combination of high surface reflectance, aerosol, and low columnar water vapor – Developed several tests to reduce these false negatives: typically gives more high-AOD events False positives also decreased through refinement of other cloud tests and QA flags 01/07/2003 Before After

RR1.38/0.66 AOT C5.1 AOT C6 CALIOP TAB, 532 nm Thin Cirrus Under-Screening over Southeast Asia, February 18 th 2002 Converse of previous case: undetected cirrus, visible in CALIOP, led to contamination (high biases) in C5.1 retrieval New reflectance ratio and tests identify this, removing the retrievals from Collection 6 Not found to introduce significant false negatives See poster here by Jaehwa Lee et al

Merged dataset C6 will include a new SDS of merged 550 nm AOD from the Deep Blue and Dark Target algorithms. To minimize pixel-level discontinuities and for simplicity/clarity, pixels will be assigned to either algorithm based on climatology of atmospherically corrected NDVI – Only Deep Blue is available for bright barren surfaces – Dark Target has longer heritage for vegetated surfaces – Note SDS will also include the ocean algorithm retrievals – Will be an interim ‘transition zone’ where retrievals will either be averaged (if the same QA) or that with higher QA flag chosen Example shown to the right for June Dark Target Deep Blue Merging zone Multiannual mean NDVI, June

Validation Validated Aqua data against AERONET – See also poster by A. M. Sayer et al., and a paper has been submitted to JGR One-sigma absolute uncertainty estimates provided for each pixel, dependent on viewing geometry and AOD For typical geometries, expected error (EE) approximately % Similar at other wavelengths (412 nm, 470 nm, 670 nm)

Correlations are high: capture variability at individual sites well Most sites have 68% or more of matchups within expected error, and small biases Performance poorer for spatially heterogeneous sites, and complex aerosol mixtures (typically parts of Africa and SE Asia)

Comparison to C5 Have validated C5 and C6 against AERONET for 10 sites where both perform retrievals C6 has a better data volume (~ double the number of matchups), better correlation with AERONET (0.93 vs. 0.86), and smaller errors on retrieved AOD (bias ~halved, RMS error decrease by ~30%)

Effect of L1b changes Ran C6 algorithm through C5 and C6 L1bs for long-term AERONET sites, to evaluate effect of radiometric calibration change – Note 2002 and 2011/2012 are incomplete due to Aqua and AERONET respectively – Differences become stronger in later years C6 L1bs result in better retrievals – Effect is smaller than that of algorithmic improvements Small trend in biases in retrievals for both calibration coefficients – Unknown whether this is coincidental, algorithmic, or calibration-related Thank you MCST!

Summary Quality: – The C6 Deep Blue dataset will have better spatial coverage (including vegetated land surfaces) and smaller uncertainties than C5 – We will be providing uncertainty estimates on an individual-retrieval basis, to aid quantitative analysis and facilitate applications such as data assimilation Status: – Aqua is ready to go – Terra testing is underway; the delay is due to the extra calibration efforts required for Terra – Algorithm and validation papers have been prepared Thanks to MCST and the Ocean Colour group for their extensive efforts in maintaining the high radiometric quality of MODIS data, and to the AERONET PIs for creation and stewardship of this invaluable resource for validation